Modelling traffic noise in a wide gradient interval using artificial neural networks.

Area of Science:

• Environmental acoustics
• Transportation engineering
• Artificial intelligence

Background:

• Classical traffic noise prediction models often overlook the significant impact of road gradients.
• Mountainous urban environments present unique acoustic challenges due to varied topography.

Purpose of the Study:

• To investigate traffic noise characteristics across diverse road gradients in Chongqing, a mountainous city.
• To develop and validate an artificial neural network (ANN) based traffic noise prediction model incorporating longitudinal gradients.
• To compare the performance of the proposed ANN model against conventional traffic noise prediction models.

Main Methods:

• Collected field data on traffic volume, heavy-vehicle ratio, average speed, road gradient, and equivalent sound pressure levels.
• Developed and optimized an artificial neural network (ANN) model, specifically a one-hidden-layer architecture.
• Validated the ANN model by comparing its predictive accuracy against two established classical traffic noise models.

Main Results:

• The optimized one-hidden-layer ANN model demonstrated superior predictive performance for traffic noise in mountain cities.
• The best-performing ANN model achieved a high coefficient of determination (0.9447) and a low mean-squared error (0.2708 dBA).
• Longitudinal road gradients were confirmed as a significant factor influencing traffic noise levels and prediction accuracy.

Conclusions:

• A one-hidden-layer artificial neural network (ANN) model is effective for traffic noise prediction in complex mountainous terrains.
• Incorporating road gradient data significantly enhances the accuracy of traffic noise prediction models.
• The developed ANN model offers a more precise alternative to conventional methods for assessing traffic noise in challenging urban landscapes.